The prediction for each day is based on a year of data prior to the day. The following figures illustrate the process of inference for today's prediction.
Figure 1 shows the year of data as a plot of event peak flux versus time (top panel). The rate of flaring is modelled as a piecewise constant Poisson process, and the Bayesian blocks procedure (Scargle 1998, ApJ 504, 405-418) is applied to the data to determine the piecewise intervals, i.e. periods of time during which the rate is approximately constant. The resulting sequence of intervals and rates is shown in the bottom panel of Figure 1. The last Bayesian block describes the current rate, and the previous blocks give information about how the rate might be expected to vary.
Figure 2 shows the distribution of the peak flux of the year of events
as a differential distribution (top panel) and as a cumulative distribution
(bottom panel), in a log-log representation.
The straight line appearance of the data reflects the fact that flares
obey a power-law size distribution. A power-law model (with a slope
determined by the maximum likelihood method) is shown in each
panel by a thick line.
Based on the information on the current and prior rate obtained by
the Bayesian blocks procedure and the power-law model for the size
distribution, posterior distributions are constructed for epsilon, the
probability of at least one flare above a specified size within one
day. The sizes considered correspond to M and X class events.
The resulting distributions are shown in Figure 3.
Additional information:
Summary of all predictions made to date
Automatic check of the reliability of the predictions
Prediction for 4 November 2003
A test of the method on historical data
Links to other pages related to flare prediction
Mike Wheatland's home page
Acknowledgement: The predictions given here are based on information from the Space Environment Center, Boulder, CO, National Oceanic and Atmospheric Administration (NOAA), US Dept. of Commerce.